Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Are Cross-wind compensation and Road crown compensation functions inputs to LKA function?

Status
Not open for further replies.

jani12

Member
Please consider Active Steering Assist Electronic Control Unit (Steering ECU) used in a commercial truck. It has Lane Keep Assist (LKA) function, Cross-wind compensation, and Road crown compensation.

This ECU is composed of a camera input, and driver outputs to control the torque overlay steering gear.

Is LKA function completely independent of above two compensation functions? For example, if this LKA function is moved to a different ECU (camera ECU) in the truck, the above two compensation functions don't need to be moved to the camera ECU. The LKA function in camera ECU would calculate torque demand and send it to steering ECU over the CAN network.

I'm thinking camera ECU doesn't need above two compensation functions for LKA functionality? I think the way LKA function works is it compensates for lateral movement by analyzing video provided by the camera? It doesn't matter what disturbances are causing lateral displacement, whether it's wind or something else, the torque demand that is calculated would be output to steering gear to compensate lateral displacement.

I'm thinking above two compensation functions are input to steering assist function and are not needed for the LKA function? I think the lateral displacement gets captured by camera and torque demand that gets calculated includes any environment disturbances, such as wind? If above two compensation functions are input to torque demand, then compensation due to wind and crown road would be applied twice?

I think whether LKA function resides in steering ECU or camera ECU, it doesn't use above two compensation functions? They are for steering assist functionality?
 
The cross-wind and road crown compensation are feed-forward control, while the LKA are feed-back control. If the signals from the cross-wind and road crown compensation succeed in keeping the vehicle on the right path, there will be no error, so the LKA won't have any work to do, so there isn't really any risk that compensation will be applied twice.

I don't know exactly why the compensations are needed as well as LKA, but the compensations may be faster to act, or they may help when the cameras that feed the LKA can't work out where the vehicle is on the road.
 
Diver300, prior to asking follow-up questions, I need to ask following question:
LKA function is completely independent of cross-wind compensation function and road crown compensation function? Another words, if LKA function resides in camera ECU, this ECU doesn't need these two compensation functions for LKA to function?
 
The functions can't be completely independent as they all affect the steering. It's quite possible that the LKA would working without the compensations, but it would be likely to not work as well. It's also possible that the compensations alone would be useful, as it's the driver who ultimately controls the direction.

I have no idea what function resides in which module. On vehicles it's quite common for functions to only work when all of their input signals are present, so you can't just remove a compensation signal if you want to see if the LKA would work without it. You would have to replace the signal with one that is saying that there is no compensation, and that would need some detailed knowledge of how the signal works.
 
I'd never heard of "road crown compensation"?

Having thought about it for a few hours, the only way I can come up with of measuring that would be to use the speed/acceleration/brake and steering angle to calculate theoretical accelerations, then compare that to the output of an accelerometer.

Any difference between calculated and actual lateral force would presumably be due to road camber??

That would presumably mean all the sensors involved would be interactive with each other and cross-dependant..
 
I don't know exactly why the compensations are needed as well as LKA
The two compensation functions are implemented in steering ECU. LKA is implemented in camera ECU. I don't think compensation functions are needed by LKA because cross-wind and road crown disturbances will be picked up by the camera and input to LKA function? If the two compensation functions are also input to LKA, then compensation will be applied twice.

I think the two compensation functions are used by steering assist function?
And since LKA finds out about these disturbances from the camera or video, these two compensation functions are not input to LKA?
I have no idea what function resides in which module.
In the truck, cross-wind compensation function and road crown compensation functions resided in the steering ECU. LKA resides in the camera ECU. LKA calculates torque demand and transmits to steering ECU over CAN. And steering ECU overlays this torque by applying it to the steering gear?
On vehicles it's quite common for functions to only work when all of their input signals are present, so you can't just remove a compensation signal if you want to see if the LKA would work without it.
The two compensation functions and LKA were implemented in the steering ECU. Maybe these two compensation functions were never input to LKA. Therefore, moving LKA to camera ECU will not change anything.
 
The two compensation functions are implemented in steering ECU. LKA is implemented in camera ECU. I don't think compensation functions are needed by LKA because cross-wind and road crown disturbances will be picked up by the camera and input to LKA function? If the two compensation functions are also input to LKA, then compensation will be applied twice.
The cameras can detect position on the road. They may also be able to detect angle between the centre line of the vehicle and the centre of the road, but they almost certainly can't detect sideways accelerations. The compensations are likely to detect sideways accelerations, so should be much faster. If the system was only looking at the road position, it would be likely to be worse control.

The compensation won't be applied twice. One is a compensation for fast events, and one for slow events.

The camera system can't respond instantly to road position. It detects the road markings, and in some cases they won't be there, or they could change suddenly, like where a single line stops and a double line starts. If the camera system tried to respond quickly to that it would be uncomfortable at best.

The steering wheel angle basically sets the sideways acceleration. The integral of sideways acceleration is sideways speed, and the integral of sideways speed is sideways position. When driving it's not a good idea to have the sideways speed or sideways acceleration get too large, as that is uncomfortable, can lead to loss of control, and can lead to overshooting the desired position.

When a driver is in a lane on the motorway, he will use the lines as guidance, but he will also take account of sideways acceleration, and how fast the car is moving sideways.

The effect of considering the sideways speed and acceleration is most obvious when a lane change is required. When the decision is made to change lane, the desired road position is suddenly 4 to 5 metres to one side. The driver needs make the car move that distance, so he needs some sideways speed, but he will limit his sideways speed to maybe 1 - 2 m/s. Also to achieve 1 - 2 m/s sideways, he will need to steer to accelerate sideways, but the sideways acceleration will probably be limited to 0.1 m/s2, so the actual steering input will be tiny at motorway speeds.

In contrast, if a car hits a big puddle at speed, it will start to turn very quickly, and the driver will have put in a big steering angle very quickly, well before the position on the road has changed much at all. He will be responding to the sideways acceleration and speed. If he waited for the road position to change much, his correction would be to little and too late.

A driver won't calculate those figures, and probably won't know he is making those adjustments, just like no-one thinks how walking works when they walk. Sideways speed depends on the angle of the car to the lane, and sideways acceleration is felt by how the seat pushes against the driver, so the driver is subconsciously aware of both all the time.

My point is that it's not just a matter of controlling one variable, the road position. The sideways speed and sideways acceleration have be controlled as well.
 
The compensation won't be applied twice. One is a compensation for fast events, and one for slow events.
Assuming lane markings are clear, why is detection of sideways position by camera a slow event. And why are compensation functions fast at detecting sideways acceleration.
If the system has compensated for sideways acceleration or velocity (fast events) by calculating torque to bring vehicle back to center of lane, why is there need to compensate for slow event (positon)?
 
Assuming lane markings are clear, why is detection of sideways position by camera a slow event. And why are compensation functions fast at detecting sideways acceleration.
If the system has compensated for sideways acceleration or velocity (fast events) by calculating torque to bring vehicle back to center of lane, why is there need to compensate for slow event (positon)?
One of the problems is that the lane markings aren't always clear.

The lane markings themselves could be imperfect and not in a totally straight line, or they could be worn or dirty. That would effectively make the lane position a noisy signal.

The sideways acceleration is the second derivative of the position, in other words the rate of change of the rate of change of the road position. Calculating the acceleration from the position would lead to a lot of noise in the calculated acceleration, so it is far better to measure the sideways acceleration directly.

Steering angle is approximately proportional to sideways acceleration, adjusting the steering can directly, and quickly, control sideways acceleration.

In the short term, for time periods of a couple of seconds or less, if the compensation for sideways acceleration works, then the need for corrections based on lane position will be substantially reduced.

However, there always needs to be some longer-term correction, to allow for small errors in measurement of acceleration. Without any other correction, a small error in sideways acceleration would build up over a few seconds to give a large error. There also needs to be something to allow for curves.
 
My point is that it's not just a matter of controlling one variable, the road position. The sideways speed and sideways acceleration have be controlled as well.
If you used acceleration to correct sideways displacement, then, you are done, meaning system or driver has returned the vehicle to original position. Why would there be need to detect position when detection of acceleration already took care of lateral displacement?
 
Example: changes in side wind load require changes in steering angle offset, proportional to the wind.
Likewise changes in camber.

If you wait for the optical system to correct it, the vehicle has already started drifting sideways.

Appropriate inputs from sensor can add steering offsets immediately, before any noticeable lane drift occurs.
That makes the vehicle control much more accurate.


As an example, a really strong side wind can cause enough force just on a normal car to need the steering wheel periphery moved up to an inch or two to maintain the lane centre, as the car passes in and out of the wind shadow of other vehicles.
The effect on large vehicles is presumably vastly greater.

The vehicle does not suddenly jump sideways, such as an accelerometer would detect - the effect feels like the steering suddenly has different offsets switched in and out, as the wind load changes.

Corrective inputs for that ideally need applying before the course of the vehicle has had chance to change and made it drift within the lane.
 
If you used acceleration to correct sideways displacement, then, you are done, meaning system or driver has returned the vehicle to original position. Why would there be need to detect position when detection of acceleration already took care of lateral displacement?
On longer timescales, the road will turn and the lateral acceleration won't be measured perfectly accurately. There is also road camber, which will mean that the target lateral acceleration, as measured by conventional accelerometers, won't be zero. The camber can also change.

As the position is the second integral of acceleration, it only takes a tiny acceleration to give a large position error given enough time.
 
In contrast, if a car hits a big puddle at speed, it will start to turn very quickly, and the driver will have put in a big steering angle very quickly, well before the position on the road has changed much at all.
Instead of driver, would the system compensate by putting a big steering angle very quickly.
 
Is driving vehicle over big puddle at speed fast event? Is vehicle drifting from it's lane due to driver not paying attention a slow event?
I'm not familiar with LKA algorithms, so my answer will be a bit of a guess. However, I have worked with control systems that have fast control parts and slow control parts.

In my estimation, hitting a puddle at speed would defiantly be a fast event, as it would suddenly put a big force on the car. At 60 mph, it would only take about 100 ms to go from road to puddle for both wheels one one side of a car.

A vehicle drifting would be something that can probably be reacted to in 5 seconds or so before the vehicle left the lane, so I would call that a slow even.
 
Why camera system cannot respond instantly to road position?
I mentioned the problems before. Lane markings can change and they may not be perfectly accurate. The camera's view of the lines may be affected by rain, or wipers, or both. Any of those issues will give a noisy signal, which the car can't respond to because it will result a very uncomfortable ride and probably wear on the steering and suspension.

For example, I don't think that an imaging system could reliably work out what is the centre of the lane through this:-https://goo.gl/maps/NZus5N977tcVG7sM6 If it tried to instantly centre the car on the wider part where the arrow is in place of the white line, and then jumped back to the narrow part after the arrow, at best it would upset the passengers, at worst, loose control of the car. It's far better to have a control system that is slower, and if something else needs to be added to respond to cross winds etc., that can be added.

Elsewhere on this electronics forum you will find lots of examples of feedback where the high frequency events are handled differently. Putting a capacitor in parallel with a feedback resistor on an op-amp is one way that is done.

In many ways, the LKA systems share features with the suspension that keeps the vehicle at the correct height. The suspension has the slower control system (the springs) and the faster control system (the shock absorbers). Further refinement is added with the tyres, the suspension bushes and the seats.

Of course the car could be made to respond as fast as possible to changes in road height, with solid tyres, virtually no suspension and no seat padding, but most people don't want to drive in something that has the ride quality of a Formula 1 car or a fork lift truck.
 
Instead of driver, would the system compensate by putting a big steering angle very quickly.
I don't know how sophisticated the LKA systems and the compensations are. There has to be a physical limit where no amount of steering input, possible also differential braking, would stop the car from leaving the lane.

My point was that a driver wouldn't wait to see that the car had left the lane, he or she would respond to the sideways acceleration of the car and try to get that to a low value. The driver would also have a reasonable idea from the intensity and duration of the acceleration as to how fast the car was travelling sideways, out of lane, and would try to correct that as soon as is reasonably possible. It is only after those things have happened that the driver will look at the lane markings and correct to those.

The skills to make those estimates of acceleration, speed and position are things we learn as babies as they are needed to walk. Automatic systems need to be told what to do, and it's a very complicated subject.
 
The vehicle does not suddenly jump sideways, such as an accelerometer would detect - the effect feels like the steering suddenly has different offsets switched in and out, as the wind load changes.
Different offsets means it feels like steering system is turning the wheels but actually wind is turning the wheels?
 
Different offsets means it feels like steering system is turning the wheels but actually wind is turning the wheels?
Almost; the wind is not turning the wheels, but the vehicle is forced sideways to some extent by transverse forces.

In other words, the steering zero or "straight ahead" angle changes with the external and internal forces on the vehicle, such as wind load, centrifugal force on curves, road camber etc.

I believe it is due to the flexibility of normal pneumatic tyres, where the part in contact with the road is distorted to the side & that's continuously replaced by undistorted or less distorted tyre as the vehicle moves - which then in turn distorts due to load, allowing a continuous sideways drift dependant on the side force & rigidity of the tyre sidewalls.

eg. "Low profile" tyres being less affected as they stiffer sidewalls, but not appropriate for goods vehicles...

>google<
Found an article with some info re. the distortion effects & it is due to tyre flexing:
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top